The research described in this article extends work done on human behavior in naming concepts and objects by adapting a methodology developed at Bell Labs and applying it to a library indexing language, the Library of Congress subject headings. The study investigated the representation of users' knowledge (names of objects and concepts), database representation for similar objects and concepts, and degree of agreement among users and between users and information system. Three user groups gave names to 40 stimuli. Names generated were compared with each other and with LC subject headings. Degree of agreement was calculated using similarity measures. The analyses identified patterns of agreement and variability in naming. There was little agreement across people in the names they used to describe texts or illustrations. There was little agreement in the names people use and the names recommended for use by LC, implying that retrieval systems should do more to accommodate common naming behavior. © 1995 John Wiley & Sons, Inc.